Benchmark Almanac

VERA-MH Suicide Risk

GPT-5.6 Terra 75.0 leads since July 2026
Not in model cards

Chatbots talk with simulated patients who show different levels of suicide risk. A judge grades the full conversation against a clinical rubric.

Spring Health ranks 28 models on suicide risk. Slingshot AI independently tested ten models using two patient simulators.

Released
October 2025
Kind
Wellbeing and sycophancy
Level
Expert
Scored
Graded by a model
Answers
Open-ended
Tasks
All public
Measures
Propensity
Language
English
In and out
Text → Text
Interaction
Multi-turn
Items from
Generated
Refresh
One fixed set
Run by
Anyone
Strongest model tested
Claude Opus 5 #11 on the capability index
Last model added
August 2026
Scores refreshed
November 20th, 2026
28
Models scored on its own leaderboard

GPT-5.6 Terra leads with 75.0 among models released by August 2026

Best score per model from the benchmark's own leaderboard, read on November 20th, 2026. Where a model ran with several settings or harnesses, we keep its best.

The newest model with a score here came out in August 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 1 GPT-5.6 Terra 75.0
  2. 2 Claude Sonnet 5 72.0
  3. 3 GPT-5.6 Sol 72.0
  4. 4 GPT-5.6 Luna 71.0
  5. 5 Claude Opus 4.5 68.0
  6. 6 Claude Sonnet 4.6 68.0
  7. 7 GPT-5.5 68.0
  8. 8 Claude Opus 4.8 68.0
  9. 9 GPT-5.4 67.0
  10. 10 Claude Opus 4.6 66.0
  11. 11 Claude Opus 5 64.0
  12. 12 Claude Fable 5 61.0
  13. 13 GPT-5.2 60.0
  14. 14 Claude Opus 4.7 60.0
  15. 15 Claude Sonnet 4.5 58.0
  16. 16 Claude Sonnet 4 42.0
  17. 17 Gemini 3.7 Flash 40.0
  18. 18 Gemini 3 Flash 39.0
  19. 19 Gemini 3.6 Flash 39.0
  20. 20 Gemini 3.5 Flash-Lite 38.0
  21. 21 Grok 3 37.0
  22. 22 Gemini 3.5 Flash 37.0
  23. 23 Claude Opus 4 35.0
  24. 24 Gemini 2.5 Flash Sep 2025 35.0
  25. 25 Gemini 3.1 Pro 33.0
  26. 26 Grok 4 32.0
  27. 27 Gemini 3 Pro 32.0
  28. 28 GPT-4o Mar 2025 28.0

Source: Official VERA-MH Suicide Risk leaderboard

#ModelScoreSettingDateSource
1 GPT-5.6 Terra 75.0 Jul 2026 Official leaderboard
2 Claude Sonnet 5 72.0 Jun 2026 Official leaderboard
3 GPT-5.6 Sol 72.0 Jul 2026 Official leaderboard
4 GPT-5.6 Luna 71.0 Jul 2026 Official leaderboard
5 Claude Opus 4.5 68.0 Nov 2025 Official leaderboard
6 Claude Sonnet 4.6 68.0 Feb 2026 Official leaderboard
7 GPT-5.5 68.0 Apr 2026 Official leaderboard
8 Claude Opus 4.8 68.0 May 2026 Official leaderboard
9 GPT-5.4 67.0 Mar 2026 Official leaderboard
10 Claude Opus 4.6 66.0 Feb 2026 Official leaderboard
11 Claude Opus 5 64.0 Jul 2026 Official leaderboard
12 Claude Fable 5 61.0 Jun 2026 Official leaderboard
13 GPT-5.2 60.0 Dec 2025 Official leaderboard
14 Claude Opus 4.7 60.0 Apr 2026 Official leaderboard
15 Claude Sonnet 4.5 58.0 Sep 2025 Official leaderboard
16 Claude Sonnet 4 42.0 May 2025 Official leaderboard
17 Gemini 3.7 Flash 40.0 Aug 2026 Official leaderboard
18 Gemini 3 Flash 39.0 Dec 2025 Official leaderboard
19 Gemini 3.6 Flash 39.0 Jul 2026 Official leaderboard
20 Gemini 3.5 Flash-Lite 38.0 Jul 2026 Official leaderboard
21 Grok 3 37.0 Apr 2025 Official leaderboard
22 Gemini 3.5 Flash 37.0 May 2026 Official leaderboard
23 Claude Opus 4 35.0 May 2025 Official leaderboard
24 Gemini 2.5 Flash Sep 2025 35.0 Sep 2025 Official leaderboard
25 Gemini 3.1 Pro 33.0 Feb 2026 Official leaderboard
26 Grok 4 32.0 Jul 2025 Official leaderboard
27 Gemini 3 Pro 32.0 Nov 2025 Official leaderboard
28 GPT-4o Mar 2025 28.0 Mar 2025 Official leaderboard

Rankings by part

We read 5 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as Detects potential risk, Confirms risk and Guides to human care. GPT-4o leads 1 of them, and 4 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.

Detects potential risk

28 models we track have a score here.

  1. 1 GPT-4o Mar 2025 ↑27 97.0
  2. 2 Claude Sonnet 5 97.0
  3. 3 Gemini 3.7 Flash ↑14 97.0
  4. 4 Claude Opus 4 ↑19 96.0
  5. 5 Claude Sonnet 4.6 ↑1 96.0
  6. 6 Gemini 3.5 Flash ↑16 96.0
  7. 7 Claude Fable 5 ↑5 96.0
  8. 8 Gemini 3.6 Flash ↑11 96.0
  9. 9 Claude Sonnet 4.5 ↑6 95.0
  10. 10 Gemini 3 Pro ↑17 94.0

4 models have held number one on the official leaderboard since its release in October 2025

Lab claims count from the day the card came out, independent runs from the model’s release.

Anthropic OpenAI
Own leaderboard Not tested since Aug 2026

Source: Official VERA-MH Suicide Risk leaderboard

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

Plot with other benchmarks
On the benchmark’s own leaderboard
20.0 40.0 60.0 80.0 2025 2026 20.0 40.0 60.0 80.0 2025 2026

Source: Official VERA-MH Suicide Risk leaderboard

No lab model card reports it yet

We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.

More in Safety and alignment

All 636 in Safety and alignment
  • FORTRESS In use Paired adversarial and benign national-security misuse prompts. Rubric-graded risk score on adversarial and over-refusal score on benign prompts. Claude Opus 5.5 5.78%
  • StrongREJECT In use Forbidden-prompt jailbreak benchmark with a rubric grader that scores how useful a non-refusal actually is. Inkling 98.6%
  • Gray Swan IPI Arena benchmark In use Indirect prompt-injection attacks from public red-teaming competitions replayed on tool, coding and browser agents. Attack success rate at k attempts. Gemini 4 Carbon 0%
  • Covert misbehavior (Thinking Machines) In use Agentic tasks that check whether a model misbehaves and conceals it. Scored by the rate of covert misbehavior. Inkling 2 Mini 8.30%
  • Destructive actions (Thinking Machines) In use Agentic tasks that check whether a model takes destructive actions it was not authorized to take. Scored by the rate of destructive actions. Inkling 2 Small 4.70%
  • Misbehavior (Thinking Machines) In use Agentic tasks that check whether a model fabricates results, ignores instructions or hides what it did when a task is hard. Scored by the rate of misbehavior. Inkling 2 Mini 9.80%

Data audit

All 9 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 3 pass 2 notes
  • Pass Independent runs Scores from the official leaderboard (28 entries).
  • Pass Models scored 28 models we track.
  • Pass Newest score The newest is from August 2026, for Gemini 3.7 Flash.
  • Note Frontier models None of the top 10 has a score. The strongest tested is Claude Opus 5 (#11).
  • Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
  • Pass Importer We read the board on vera-mh.com: 28 rows for models we track at the last read.
  • Pass Last read November 20th, 2026.
  • Pass Read worked The last read worked.
Consistency 1 pass 1 note
  • Note Lab claims No model has both a lab number and an independent run.
  • Pass Known issues No published audit found problems.

How we check